A Demonstration Application

We used the results of this Chapter to implement an application to evaluate video quality in real time for all the ranges of the chosen quality-affecting parameters (namely: LR, FR, CLP, BR, and RA). We have extended the simulator described in Section 6.2 with a graphical user interface to control the variation of the input parameters and to visualize the video under test. (Note that the simulator already supports the packetization and network transmission simulation.) The application incorporates two operating modes: manual and automatic modes as depicted in Figures 6.14 and 6.17 respectively. In the manual mode, the five parameters can be changed manually. The network parameters (LR and CLP) can be varied smoothly during video playback. However, the codec's parameters (BR, FR, and RA) have to be fixed before starting playback of the video sequence under test. This is because the encoder does not support dynamic variation of the encoding parameters. In addition, we have only considered CBR video transmission. The parameters collector module probe the values of the parameters regularly, every predefined interval (currently fixed to 300 ms). The trained NN module (currently RNN is used) measures the quality based on the values of the parameters. In the automatic mode, the application plays back the video sequences and distorts them from a predefined sets of parameters' values. The goal of this mode is to study the variation of the quality with the parameters (varying one of them and keeping the others fixed). In the current implementation, we integrated three standard video sequences (namely Stefan, Children, and Foreman).. Each of these sequences has different video characteristics and nature. ``Stefan'' is characterized by fast motion, while the ``Children'' sequence is characterized by slow motion, more spatial redundancy and more colors saturation. The ``Foreman'' sequence is a moderate one between ``Children'' and ``Stefan'' sequences. It should be noted that the BR of the best quality of the encoded versions are 1345, 1200, 1000 KByte/s for Stefan, Foreman, and Children respectively. By using the algorithm described in Section 6.3, our application work quite well for all the three video sequences regardless of the variability of the BR from sequence to another. We show the manual mode with the tested video sequences for Stefan, Children, and Foreman in Figures 6.14, 6.15, and 6.16 respectively.

Figure 6.14: A screen dump showing manual mode for Stefan

Figure 6.15: A screen dump showing manual mode for Children

Figure 6.16: A screen dump showing manual mode for Foreman

Figure 6.17: A screen dump showing Automatic Mode for Stefan